Let me raise a skeptical eyebrow.
True, unmanned aerial vehicles—some call them "drones"—are increasingly roaming the skies over Afghanistan and Pakistan, armed with high-res cameras and extremely accurate missiles, guided, aimed, and fired by "joystick pilots" watching screens and pushing buttons from "ground control stations" back at Creech Air Force Base, Nev.
Last year, for the first time, the U.S. Air Force trained more personnel to man these remote-control joysticks than to pilot combat aircraft from inside the cockpits.
The Army has its swarms of mini-cams, too, attached to the tip of projectiles, peeking around corners or inside buildings to see whether enemy combatants are lurking. They also have R2D2s sniffing out, and disarming, roadside bombs.
But these devices have changed only the face of modern warfare, not its nature. They're tactical tools, not strategic transformers. They are firmly under human control. And to the extent that the U.S. military continues to focus more on counterinsurgency than on traditional modes of combat, this will remain the case emphatically.
A little historical background. The concept for these high-tech weapons was hatched in the mid-1970s, when the Army and the Defense Advanced Research Projects Agency were scanning the drawing boards for new technologies that might help NATO stave off a Soviet invasion of Western Europe. A defense guru named Albert Wohlstetter, working on contract for DARPA, found two promising possibilities: a pair of "remotely piloted vehicles"—at the time, primitive model airplanes powered by modified lawn mower engines, able to loiter for two hours while carrying a 28-pound payload—and a bunch of new sensors, based on lasers, millimeter waves, or the signals from the (then-fledgling) global positioning satellite system that could guide a missile to within a few feet of its target.
Wohlstetter came up with the idea of fusing the two projects together. In a (since-declassified) report written in 1975, he proposed putting a camera inside the belly of the RPV; it would scan the ground along the flight path and transmit the images back to a base, where an officer would steer the vehicle toward the target by remote control. The RPV could also carry a bomb—fired by the same remote officer—that would be guided to the target by the new sensors.
This idea became the Army's Assault Breaker program and, over the next quarter-century, the basis for several Army, Navy, and Air Force systems, including the drones flying over Iraq, Afghanistan, Pakistan, and elsewhere today.
The end of the Cold War roughly coincided with Saddam Hussein's invasion of Kuwait—and gave U.S. Air Force strategists a more ambitious idea: that these new weapons might revive the age-old dream of a war fought and won almost entirely from the air.
Two Air Force colonels, David Deptula and John Warden, devised a list of 84 targets in Iraq, mostly in Baghdad—the "key nodes" that held together Saddam's military command. Destroy those nodes (and this could be done with the new "smart bombs," as the accurate munitions were called), and Saddam's regime would collapse like the proverbial house of cards.
The targets were destroyed, but the regime didn't fold.
At the time, we didn't have many of these smart bombs; they were expensive ($250,000 apiece), and they weren't so smart. They were guided by laser beams, which were refracted or dissipated by smoke and dust, causing the weapon to veer off.* Through the monthlong air campaign of Desert Storm, only 9 percent of the bombs dropped were smart bombs. Toward the end, B-52s flew in and carpet-bombed Iraqi positions in the field with old-fashioned dumb bombs. And even then, the war wasn't won until about 500,000 ground troops, who had been mobilizing for months, pushed the Iraqi army back across the border.
By the end of the 1990s, some truly smart bombs came into the arsenal. They were called JDAMs (for Joint Direct Attack Munitions); they came in kits that could be attached to the fins of any bomb in the fleet; they cost only $25,000 apiece; they were guided to their targets by GPS signals (impervious to smoke and weather)—and they were joined with a new generation of RPVs called Predator drones.
Wohlstetter's vision was realized.
We have all since heard the story of the special-ops officer, riding horseback outside the Afghan town of Mazar-i-Sharif, on Oct. 15, 2001. Through his night-vision binoculars, he spotted a regiment of Taliban fighters across the way. He pulled out a laptop computer, typed the enemy's coordinates, and pushed the "send" button. A Predator drone, hovering 20,000 feet overhead, received the message and beamed it to Prince Sultan Air Force Base in Saudi Arabia. There, a U.S. officer sent back a signal, directing the Predator to fly over the Taliban. A video camera in the drone's belly streamed the image back to the base. The officer then ordered a B-52 pilot, patrolling nearby, to attack the target. En route, the pilot punched the target's coordinates into the GPS receiver of one of his JDAMs. He fired the JDAM, which darted toward the target, exploded, and killed the Taliban.
The time that elapsed—from the special-ops officer punching in the data to the pilot dropping the bomb—was 19 minutes. A decade earlier, the sequence would have taken three days. A few years before that, it could not have occurred at all.
Over the next few weeks, the sequence was replicated with variations all across Afghanistan—phenomenally accurate airstrikes, followed by offensives on the ground by anti-Taliban insurgents, assisted by small teams of U.S. soldiers, Marines, Green Berets, or CIA advisers.
Five weeks after the war began, the Taliban fled Kabul and U.S. allies moved in. By the end of the year, a new interim government led by Hamid Karzai took office with international support.
Defense Secretary Donald Rumsfeld proclaimed the campaign a vindication of the "revolution in military affairs" and "transformation."
But he spoke too quickly. First, as we have since learned all too well, the Taliban were not vanquished. And as we departed for Iraq, they gradually returned. It's not that we won the battle but didn't win the war; it's that we didn't finish the war.
Second, and more to the point, the wonder weapons did not win the battles, much less the war, all by themselves. After the first few JDAM strikes, the Taliban learned how to take shelter and apply camouflage so the drones' cameras couldn't see them. Some of the fiercest firefights came in the spring of 2002, months after Rumsfeld had declared victory and moved on.
The JDAMs, in retrospect, played the same role that artillery strikes and combat air support had always played—true, with much greater accuracy and speed. But it was still ground troops—human beings wearing boots on the ground—that had to go in and close the battle.
What has happened, in fact, is not so much a revolution in warfare as a revolution in the U.S. Air Force. Far from fulfilling the dream of wars waged far above the crude skirmish of terrestrial battle, the age of the drones has brought back the days when the chief mission of the Air Force was to support troops on the ground.
Nor are these weapons in any sense "autonomous robots," as Allenby puts it. In fact, the Air Force has gone back to calling them "remotely piloted vehicles" because they've discovered that these systems require far more personnel than they'd anticipated.
There's no pilot in the plane, but that's the only unmanned piece of the loop. Right now, 108 Predators, Reapers, and Global Hawks—the RPVs that do most of the work—are flying 42 combat air patrols (known as CAPs) around the clock. (By the end of next year, they'll do 50; by the end of 2013, they'll do 65.) Three RPVs are required for each CAP: one to loiter over the area of combat, one to fly toward the area, and one to fly back to a maintenance depot.
Back at the ground-control station at Creech, Nev., each CAP requires 43 military personnel rotating in three shifts. They include seven pilots, seven system operators, and five mission coordinators. Their work is backed by an intelligence unit at CIA headquarters in Langley, Va., consisting of (again, for each CAP) 66 people, including 34 video crewmembers and 18 intelligence analysts. (These numbers come from Air Force public affairs.)
The day may come when a machine analyzes the video images, distinguishes friend from foe, decides when to shoot, and pulls the trigger. But that day is not here by any stretch of the imagination. And trends in U.S. Army and Marine doctrine are pushing that day farther into the future.
Counterinsurgency campaigns, such as those in Afghanistan (and, since 2007, in Iraq), are essentially about ground troops who don't just fight insurgents but get to know the local leaders, cultivate intelligence sources, find out people's needs, then (at least in theory) help to provide services through joint civilian-military teams.
Gen. Stanley McChrystal, the U.S. commander in Afghanistan, has laid down new ground rules designed to minimize accidental shootings of civilians, even to the point of making his troops take greater risks. This is about strategy, not morality: The more civilians our troops kill (however understandable the mistake), the more we inflame anti-American sentiments and swell the ranks of the insurgents and their supporters.
Yet if troops on the ground (or in helicopters) occasionally mistake civilians for insurgents, robots in the sky (or elsewhere) would make these mistakes far more often.
Singer and Allenby, of course, warn of this danger. But they exaggerate its imminence.
Correction, May 19, 2010:
This piece originally and incorrectly said that the smart
bombs of the early 1990s were guided by radar. (Return
to the corrected sentence.)